A Novel Fluorescent Protein-Based Biosensor for Gram-Negative Bacteria

Abstract

Site-directed mutagenesis of enhanced green fluorescent protein (EGFP) based on rational computational design was performed to create a fluorescence-based biosensor for endotoxin and gram-negative bacteria. EGFP mutants (EGFP_i) bearing one (G10) or two (G12) strands of endotoxin binding motifs were constructed and expressed in an Escherichia coli host. The EGFP_i proteins were purified and tested for their efficacy as a novel fluorescent biosensor. After efficient removal of lipopolysaccharide from the E. coli lysates, the binding affinities of the EGFP_i G10 and G12 to lipid A were established. The K_D values of 7.16 × 10⁻⁷ M for G10 and 8.15 × 10⁻⁸ M for G12 were achieved. With high affinity being maintained over a wide range of pH and ionic strength, the binding of lipid A/lipopolysaccharide to the EGFP_i biosensors could be measured as a concentration-dependent fluorescence quenching of the EGFP mutants. The EGFP_i specifically tagged gram-negative bacteria like E. coli and Pseudomonas aeruginosa, as well as other gram-negative bacteria in contaminated water sampled from the environment. This dual function of the EGFP_i in detecting both free endotoxin and live gram-negative bacteria forms the basis of the development of a novel fluorescent biosensor.

Increasing concern regarding the microbiological safety of food, water, dairy products, industrial waste, and pharmaceutical preparations has provided an urgency for detection methods that are fast, sensitive, specific, reliable, and quantitative for quality assurance in order to prevent infections and epidemics (10). There are a large number of detection methods for microorganisms, including immunomagnetic separation and flow cytometry (18), flexural plate wave (16), quartz crystal (8), and surface acoustic wave (9). In addition, due to the ubiquity and lethality of endotoxin (or lipopolysaccharide) from the outer cell-wall of gram-negative bacteria, many pharmaceutical products are rigorously tested for the presence of contaminating lipopolysaccharide or gram-negative bacteria before the products are sold for human use. Both the United States Pharmacopoeia and the European Pharmacopoeia specify the rabbit pyrogen test and the Limulus amoebocyte lysate test as the quality control tests for the presence of endotoxin in injectables and medical devices.

A number of new approaches to pyrogen testing have been reported. These are mainly based on an in vitro pyrogen test involving the use of human cells such as leukocyte cell lines, isolated primary blood, and whole blood (4, 20). Recently, genetic engineering of an endotoxin-sensitive Limulus amoebocyte lysate protein, recombinant factor C expressed in a baculovirus system, produced an enzyme with remarkable sensitivity to endotoxin, at 0.001 endotoxin unit (EU)/ml (2).

The green fluorescent protein (GFP) has been a popular choice for development of reporter-biosensors to detect various environmentally hazardous compounds (7, 11, 19). Through computer-aided simulation and rational design, we have recently developed a fluorescent biosensor for lipopolysaccharide and lipid A (the bioactive moiety of lipopolysaccharide) with enhanced green fluorescent protein (EGFP) as a scaffold protein (6). Previously, we have shown (5) that lipopolysaccharide or lipid A can interact strongly with short cationic amphipathic sequences of five alternating basic (B) and hydrophobic (H) residues (BHBHB). Thus, such sequence motifs were introduced into the β-sheets located on the surface of the EGFP barrel in the vicinity of the chromophore (6). The EGFP mutants (EGFP_i) showed a range of lipid A binding affinities (26.12 to 0.13 μM), resulting in concentration-dependent fluorescence quenching (6).

The high level of endogenous lipopolysaccharide, which is the ligand that binds EGFP_i, and a serious host incompatibility problem that may result in its growth inhibition or even cell death during the expression of recombinant EGFP_i in Escherichia coli are the major challenges facing the expression of EGFP_i in a gram-negative bacterial host. However, this problem did not arise in the present study. Furthermore, Schnaitman (17) has demonstrated that treatment of E. coli with the combination of Triton X-114, EDTA, and lysozyme resulted in solubilization of all lipopolysaccharide from the cell wall. Hence, our strategy to overcome lipopolysaccharide contamination of EGFP_i proteins was to target recombinant EGFP_i either into insoluble inclusion bodies or into the periplasmic space or to efficiently remove lipopolysaccharide from soluble cytoplasmic EGFP_i after lysing the cells.

We report here the construction of EGFP_i, expression in E. coli, purification, and lipopolysaccharide removal, followed by characterization and the use of EGFP_i as a novel biosensor for detection of endotoxin and live gram-negative bacteria. We tested the binding affinity of EGFP_i to lipid A over a wide range of pHs and ionic strengths. In addition, we demonstrated fluorescence quenching of EGFP_i upon interaction with lipopolysaccharide and lipid A. The fluorescent EGFP_i was used to detect gram-negative bacteria directly in laboratory cultures as well as in environmental water samples.

MATERIALS AND METHODS

Construction of EGFP_i plasmid.

After insertion of EGFP into the pBluescript SK II vector (Stratagene, La Jolla, Calif.) to yield pBS-EGFP, a lipid A binding motif was introduced into the β-sheet(s) of the EGFP scaffold via site-directed mutagenesis (6). The full-length EGFP_i mutants were synthesized by PCR with forward primer 5′-CCGCCCATATGGTGAGCAAGGGCG-3′, containing an NdeI site, and reverse primer 5′-GGGGATCCCGCGGGCCCTCTAGACT-3′, containing a BamHI site. For directional cloning, the PCR products were cleaved at the NdeI and BamHI sites and cloned into the pET3b (Novagen, Madison, Wis.) vector linearized with compatible restriction sites. With this strategy, the single- and double-motif mutants, collectively referred to as pET3b-EGFP_i, were transformed into E. coli TOP10 competent cells (Invitrogen, Carlsbad, Calif.), from which plasmid DNAs were extracted for verifying the DNA sequences before their transformation into the expression host, E. coli BL21(DE3) (Novagen, Madison, Wis.) for protein production.

Expression of EGFP_i.

The EGFP_i proteins were expressed optimally in 5 ml of Luria-Bertani (LB) medium containing 80 μg of ampicillin per ml and incubated at 37°C without isopropylthiogalactopyranoside (IPTG) induction for 16 h with constant shaking at 230 rpm. The culture was subsequently scaled up to 200-ml volume for the production of EGFP_i under the same conditions. The cultures were pelleted at 5,000 × g for 10 min at 4°C and resuspended with 40 ml of lysis buffer containing 10 mM Tris-Cl, pH 7.5. The bacterial cells in the suspension were passed through a French press (Basic Z model; Constant System, Warwick, United Kingdom) at 100 MPa of pressure for four rounds in order to generate >90% cell disruption.

Purification of EGFP_i proteins.

Soluble EGFP_i proteins were subjected to organic extraction (22). Briefly, insoluble material in the disrupted cell suspension which did not display green fluorescence was first removed by centrifugation at 20,000 × g for 30 min at 4°C. Triethanolamine base (Sigma, St. Louis, Mo.) and ammonium sulfate were added to the fluorescent green supernatant to final concentrations of 100 mM and 1.6 M, respectively. After incubation on ice for 1 h, the precipitated proteins were removed by centrifugation at 5,000 × g for 20 min at 4°C. Ammonium sulfate was added to the supernatant at room temperature to a final concentration of 2.8 M to achieve 70% saturation. The entire suspension was extracted twice by vigorous shaking for 1 min each with 1/4 (vol/vol) followed by 1/16 (vol/vol) ethanol. The aqueous and ethanol phases were separated by centrifugation at 3,000 × g for 5 min at room temperature. A 1/4 (vol/vol) concentration of n-butanol was added to the combined ethanol extract. After vigorous shaking for 30 s, the phases formed were separated by centrifugation as before. At this step, EGFP_i moved almost completely into the lower aqueous phase. The upper organic phase was discarded, and an equal volume of chloroform was added to the aqueous phase. After extraction for 30 s, the phases were separated by centrifugation as before. The upper aqueous phase containing EGFP_i was collected and dialyzed twice with 200 ml each of 10 mM Tris-HCl, pH 7.5, for 4 h with snakeskin (Pierce, Rockford, Ill.) dialysis tubing with a 3,500-Da cutoff.

Lipopolysaccharide removal with Triton X-114 and affinity chromatography.

The dialyzed EGFP_i extracts were subjected to various methods of lipopolysaccharide removal before functional studies. Triton X-114 was added to the protein preparation to a final concentration of 1%. The mixture was incubated at 4°C for 30 min with constant stirring to ensure a homogenous solution (12). The sample was then incubated at 37°C for 10 min and centrifuged at 20,000 × g for 10 min at 25°C. The upper aqueous phase containing the protein was carefully removed and subjected to Triton X-114 phase separation for three more cycles.

To further remove endotoxin, 6 to 8 ml of EGFP_i extracts was passed through 1 ml of Detoxi-Gel endotoxin-removing resin, prepacked in a 5-ml disposable column (Pierce, Rockford, Ill.) by gravity. The column was washed once with 5 ml of 1% sodium deoxycholic acid (Sigma), followed by 5 ml of 2 M NaCl, and thrice with 5 ml each of pyrogen-free water before and after each lipopolysaccharide removal. Under pyrogen-free conditions, the EGFP_i extracts were further chromatographed through a column of 1 ml of S3Δ peptide affinity gel (3) to further remove traces of endotoxin, following the same steps as for the Detoxi-Gel endotoxin-removing column. The resulting EGFP_i was assayed for endotoxin by the Limulus amoebocyte lysate chromogenic assay with Kinetic-QCL (BioWhittaker Inc., Walkersville, Md.).

Protein quantification and determination of expression level of EGFP_i.

The total protein of all cell lysates and extracts obtained from various steps of purification was quantified by the method of Bradford (1). To determine the expression level of the EGFP_i mutant proteins in the lysates and insoluble fractions, the proteins were electrophoresed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and the relative amount of EGFP_i in the bands was quantified densitometrically with Image Master VDS software (Amersham Biosciences, Buckinghamshire, United Kingdom).

Measurement of binding affinity of EGFP_i with lipid A under different conditions.

The surface plasmon resonance sensorgrams were recorded to determine the association and dissociation kinetics of EGFP_i to the immobilized lipid A with a BIACORE 2000 instrument (Biacore, Uppsala, Sweden). Briefly, lipid A (f583, E. coli; Sigma) at 1 mg/ml (0.5 mM) in water was immobilized on the HPA (hydrophobic) chip according to the manufacturer's specifications. Stock solutions of EGFP_i purified by affinity chromatography through an S3Δ column were diluted with different buffers and injected at five different concentrations (0.5 to 6 μM) over the monolayer of immobilized lipid A at a flow rate of 20 μl/min, with the diluent as the running buffer. Purified recombinant GFP (Clontech, Palo Alto, Calif.) was injected as the negative control. The dissociation constant of each lipid A-EGFP_i complex was calculated with the BiaEvaluation software, version 3.0 (Biacore). To regenerate the HPA chip, 100 mM NaOH was injected until the response unit (RU) returned to baseline. S3Δ (NH₂-HAEHKVKIKVKQKYGQFPQGTEVTYTCSGNYFLM-COOH), which binds lipid A at high affinity (21), was used as a control.

Fluorescence measurements of complexes of lipopolysaccharide-EGFP_i and lipid A-EGFP_i.

To assess changes in the fluorescence of the EGFP_i after interaction with lipopolysaccharide and lipid A, fluorimetric assays were carried out. Ten picomoles (270 ng) of G10 (or G12) sampled from three stages of lipopolysaccharide removal (dialysis, Triton X-114 treatment, and S3Δ affinity chromatography) was diluted in 100 μl of pyrogen-free buffer containing 50 mM Tris-Cl (pH 7.3) and mixed with 2 μl of various amounts of lipid A at 2.5 to 80 ng (1.25 to 40 pmol) or lipopolysaccharide at 7.5 to 240 ng (1.25 to 40 pmol). Lipid A and lipopolysaccharide were warmed at 37°C, sonicated for 30 min, and vortexed again immediately before addition to the EGFP_i extracts.

The fluorescence spectra were measured with an LS-50B spectrofluorimeter (Perkin Elmer, Beaconsfield, United Kingdom). In the recorded emission scans, emission intensities were monitored from 370 to 600 nm, while the excitation wavelength was fixed at 488 nm. The scanning speed was fixed at 1,500 nm/min, and the emission wavelength window was set at 10 nm. Changes in the emitted fluorescent light intensity at 508 nm of the lipid A-EGFP_i and lipopolysaccharide-EGFP_i complexes due to increasing amounts of lipid A or lipopolysaccharide were recorded. Native EGFP extract was used as a control to normalize the changes of fluorescence to G10 and G12.

Fluorescent tagging of live bacteria.

Laboratory cultures of Pseudomonas aeruginosa ATCC 27853 and E. coli TOP10 were challenged with EGFP_i. Briefly, 1 ml of overnight cultures containing 10¹⁰ CFU/ml was pelleted and resuspended in 1 ml of 0.9% saline. Aliquots (20 μl) of the bacterial suspension were mixed with 5 μl of EGFP_i in a final concentration of 2.0 μM. The reactions were stopped at 3, 5, and 10 min by pelleting the bacteria, washing the pellets three times with 200 μl of saline, and finally resuspending the cells in 20 μl of 0.9% saline. Bacteria tagged with EGFP_i were viewed with a FluoView 300 confocal laser scanning microscope with an IX70 inverted microscope (Olympus). To test for the specificity of EGFP_i for gram-negative bacteria, other microorganisms such as the gram-positive bacterium Staphylococcus aureus ATCC 25923 and the yeast Pichia pastoris were treated the same way. In principle, lack of lipopolysaccharide on the gram-positive bacterium and yeast strains would not result in EGFP_i tagging on these microorganisms.

Testing of contaminated environmental water samples.

Water samples collected from a fish aquarium, drain water, and a stagnant pool were centrifuged at 10,000 × g for 5 min at room temperature. From each milliliter of sample collected, the pellet was resuspended in 20 μl of saline, except for the fish aquarium water, where 5 ml of sample was pelleted and resuspended into 20 μl due to anticipated low cell density. EGFP_i at a 2 μM final concentration was added to the mixture for 10 min, and the sample was pelleted as before, washed three times with 200 μl of saline, and finally resuspended in 20 μl of 0.9% saline for observation under the confocal microscope. In parallel, these water samples were plated on Mueller-Hinton II agar (Becton Dickinson, Cockeysville, Md.) plates and incubated at 37°C overnight. Gram staining was carried out on both the water samples and the cultures.

RESULTS

Expression of EGFP_i mutant proteins in E. coli.

EGFP_i G10 and G12, bearing one and two strands, respectively, of the endotoxin binding motif(s), containing symmetrical sequences of alternating basic and hydrophobic (BHBHB) residues on one and two adjacent β-sheets were constructed and expressed in a bacterial host. The expression used the advantageous properties of EGFP: fluorescent, highly soluble, and of low toxicity. Furthermore, it requires no posttranslational modifications, which is highly in contrast to the potentially toxic effect of lipopolysaccharide-binding cationic peptides. Clones harboring the pET3b-EGFP (native) and pET3b-EGFP_i mutants (G10 or G12) constitutively expressed native EGFP and mutant EGFP_i intracellularly at 3 to 8% of total cellular protein (Fig. 1A). Consistent with expression profiles in COS-1 cells (6), the native EGFP and mutant G10 were expressed mainly in the E. coli soluble lysates, while most of G12 was located in the insoluble inclusion bodies (Fig. 1B). Other constructs were generated to target the EGFP_i to the periplasm or to carry a His tag. However, the yield was low, and most of the mutant proteins were probably misfolded and hence showed low or no fluorescence (data not shown).

FIG. 1.

(A) Coomassie blue-stained SDS-PAGE gel of EGFP_i in whole-cell (W), soluble lysate (L), insoluble fractions (I), dialyzed organic extract (D), and after S3Δ column lipopolysaccharide removal (S3Δ). G12 showed the lowest expression level at 3% of total protein, while native EGFP (G0) and G10 represented 7 to 8% of total protein. The majority of G12 was insoluble, and hence it was not purified by the organic extraction method. (B) Western analysis of EGFP_i in whole-cell, soluble lysate, and insoluble fractions. Most of the expressed native EGFP (G0) and G10 were in the soluble lysate, while G12 was mostly in the insoluble pellet.

EGFP_i protein extraction and lipopolysaccharide removal.

After organic extraction of the bacterial lysates, the enriched soluble EGFP_i yielded green fluorescence except for G12, which contained much less soluble protein. Since G12 was expressed mainly as insoluble inclusion bodies, only minimal levels of soluble protein were purified. The majority of G12 would have to be denatured and renatured from the inclusion bodies.

Stepwise treatment with Triton X-114 successfully removed the bulk of lipopolysaccharide, while the Detoxi-Gel endotoxin-removing column further purified the EGFP_i. The recalcitrant problematic trace levels of lipopolysaccharide remaining with EGFP_i were removed with an S3Δ affinity column, resulting in virtually pyrogen-free EGFP_i containing <0.05 EU of lipopolysaccharide per ml (Table 1).

TABLE 1.

Purification and LPS removal of the expression of EGFP_i

Purification stepa	Protein			EGFPi				Lipopolysaccharide removal
	Vol (ml)	Protein concnb (mg/ml)	Total protein (mg)	Protein/protein ratioc (μg/μg of total protein)	Amt (mg)	Recovery (%)	Purification (fold)	Concnd (EU/ml)	Efficiency (fold)
EGFPi G0
Crude lysate	67	24.29	1,627.43	0.035	56.96	100	1	—e	—
After centrifugation (30,000 × g)	65	17.05	1,108.25	0.021	23.27	40.85	0.6	>5,000	1
Organic extraction	19.5	2.14	41.73	0.53	22.12	38.83	15.14	NAf	NA
Dialysis	22	1.67	36.74	0.54	19.84	34.83	15.43	400	>12.5
Triton X-114	15	1.84	27.6	0.67	18.49	32.34	19.14	0.6	>8,000
Detoxi-Gel chromatography	12	2.4	25.68	0.71	18.23	32	20.29	0.04	>125,000
S3Δ chromatography	12	1.5	18	0.74	13.32	23.38	21.14	0.02	>250,000
EGFPi G10
Crude lysate	68	20.16	1,370.88	0.033	45.24	100	1	—e	—
After centrifugation (30,000 × g)	66	9.86	650.76	0.019	12.36	27.32	0.58	>5,000	1
Organic extraction	10.5	2.13	22.365	0.52	11.63	25.71	15.76	NA	NA
Dialysis	18	1.03	18.54	0.54	10.01	22.13	16.36	1,000	>5
Triton X-114	7	1.09	7.63	0.62	4.73	10.46	18.79	5	>1,000
Detoxi-Gel chromatography	6.5	1.02	6.63	0.68	4.51	9.97	20.61	0.07	>70,000
S3Δ chromatography	6.5	0.48	3.12	0.71	2.22	4.91	21.52	0.01	>500,000
EGFPi G12
Crude lysate	78	40.05	3,123.9	0.029	90.5931	100	1	—e	—
After centrifugation (30,000 × g)	76	8.65	657.4	0.005	3.287	7.27	0.17	>5,000	1
Organic extraction	18	1.65	29.7	0.045	1.3365	2.95	1.55	NA	NA
Dialysis	28	0.84	23.52	0.054	1.27008	2.81	1.86	200	>25
Triton X-114	14	1.22	17.08	0.074	1.26392	2.79	2.55	4	>1,250
Detoxi-Gel chromatography	14	1.04	14.56	0.076	1.10656	2.44	2.62	0.04	>125,000
S3Δ chromatography	13.5	1.01	13.635	0.08	1.0908	2.41	2.76	0.025	>200,000

^aNative EGFP (G0) and mutant EGFP_i (G10 and G12) were expressed in E. coli. G10 and G12 bear single and double strands of lipopolysaccharide/lipid A binding motif of the type BHBHB, respectively (5, 6), on the β-sheet in the vicinity of the EGFP chromophore.

^bTotal protein concentration was determined by the Bradford method (1).

^cEGFP_i concentration was determined by loading a fixed amount of total protein in SDS-PAGE gels, and bands were densitometrically analyzed with Image Master VDS software version 2.0 (Amersham Pharmacia Biotech).

^dLAL-QCL kinetic endotoxin test (BioWhittaker), the most effective and sensitive kit for detecting minute quantities (lowest limit, 0.005 EU/ml) of endotoxin in solutions, was used to quantify the efficiency of lipopolysaccharide removal.

^e—, not detected in crude lysates because they contained soluble as well as insoluble lipopolysaccharide on the cell wall.

^fNA, not applicable. Levels in organic extracts were not determined because they contain high salt concentrations, which may interfere with the enzymatic reaction.

Affinity of G10 and G12 for lipid A remains high over a wide range of pHs and salt conditions.

The binding affinity of G10 and G12 for lipid A was measured by surface plasmon resonance under various conditions of pH and ionic strength. Figure 2A shows a series of sensorgrams obtained by injecting increasing concentrations of G12. The slight increase in the initial response was partially attributed to the mass transfer of the bulk proteins in EGFP_i and the other minor molecules which may be present in the extracts. Thus, sensorgrams were obtained under various buffer conditions and salt concentrations, from which the corresponding K_D values were calculated.

FIG. 2.

(A) Surface plasmon resonance sensorgrams of G12 protein after purification through the S3Δ affinity column, injected at different concentrations (0.5 to 6 μM) into the flow cell, which was coated with lipid A (E. coli f583). The buffer was 50 mM Tris-HCl (pH 7.3) containing 25 mM NaCl. The surface plasmon resonance curves were used for calculation of K_D. (B) The effects of pH on binding affinity (K_D) of EGFP_i to lipid A. As the pH increased from 6 to 9, the K_D increased from 1.07 × 10⁻⁷ M⁻¹ to 4.68 × 10⁻⁶ M⁻¹ for G12 and from 3.94 × 10⁻⁷ M⁻¹ to 2.70 × 10⁻⁵ M⁻¹ for G10. (C) Effects of salt (0 to 200 mM NaCl) in 50 mM Tris-Cl (pH 7.3) on the binding affinity of EGFP_i to lipid A. EGFP_i is tolerant to salt. Between 0 and 200 mM NaCl, the binding affinity improved by 0.5 to 1 order of magnitude, with a K_D of 3.15 × 10⁻⁶ M⁻¹ to 7.16 × 10⁻⁷ M⁻¹ for G10 and 7.56 × 10⁻⁷ M⁻¹ to 8.18 × 10⁻⁸ M⁻¹ for G12. In contrast, the S3Δ control peptide showed a slight drop in binding affinity as the salt concentration increased. The binding affinity of native EGFP (G0), which has no specific binding site for lipopolysaccharide, remained low at all salt concentrations tested. The K_D values are means ± standard deviation of three individual experiments with five different concentrations.

It is noteworthy that in the absence of lipid A/lipopolysaccharide, at pH 6.0, the fluorescence of EGFP_i itself was reduced by 50%, and all extracts showed turbidity at pH 5.0 and precipitated out at pH 4.0 (data not shown). The fluorescence intensity of G10 and G12 decreased between pH 7.0 and 4.5 but remained stable between pH 7.0 and 11.0, as was also observed for GFP (14). Increase in pH decreases the binding affinity of G10 and G12, as reflected by the lowest K_D of 1.07 × 10⁻⁷ M at pH 6.0 to the highest K_D of 4.68 × 10⁻⁶ M at pH 9.0 for G12, and a K_D of 3.94 × 10⁻⁷ M at pH 6.0 to the highest K_D of 2.70 × 10⁻⁵ M at pH 9.0 for G10 (Fig. 2B). The apparent increase in K_D of G10 over pH 6 to 9 did not, in practice, adversely affect its binding to lipid A, as there is probably compensation by hydrophobic interaction between the lipid A acyl chains and the hydrophobic residues in the lipopolysaccharide-binding motif, BHBHB. This shows that the initial driving force for the biointeraction between EGFP_i and lipid A is via electrostatic interaction of the phosphate head groups of the lipid A and lysine residues on the endotoxin binding site(s) on EGFP_i, which presumably protrudes from the immobilized surface. The binding affinity increased by 5- to 10-fold when the salt concentration was raised from 0 to 200 mM NaCl (Fig. 2C), with optimal K_D at the physiological range. This may be the physiological condition that is appropriate for biointeraction between EGFP_i and lipid A or lipopolysaccharide.

Table 2 compares the kinetic constants and unit fluorescence of EGFP_i under different stages of lipopolysaccharide removal and different buffer conditions. In phosphate-buffered saline, the binding affinity dropped. This could be attributable to the competition between phosphate ions and lipid A phosphate head groups for the lysine residues on the binding site(s) of EGFP_i. On the other hand, the K_D value of native EGFP remained relatively unchanged at a low basal level, showing lack of affinity for lipid A under all conditions tested. Removal of endogenous lipopolysaccharide from the EGFP_i extracts resulted in enhanced binding of exogenously added lipopolysaccharide.

TABLE 2.

Kinetic constants of native EGFP and mutant EGFP_i to lipid A in different diluents and before removal of lipopolysaccharide

Diluenta	EGFPi G10		EGFPi G12		EGFP (G0)		S3Δ
	KD (10−6 M−1)	FU/μMb	KD (10−6 M−1)	FU/μM	KD (10−3 M−1)	FU/μM	KD (10−6 M−1)
Water	3.10c	410.7 ± 11.5	0.44	15.3 ± 2.5	2.59d	605.1 ± 150.3	0.59
Tris-HCl, 50 mM, pH 7.3	4.18	394.4 ± 4.2	0.90	14.7 ± 0.2	1.81	653.7 ± 3.5	0.011
150 mM NaCl in Tris-Cl, pH 7.3	9.58	426.3 ± 9.6	0.55	14.3 ± 0.5	0.10	599.6 ± 0.3	0.83
Phosphate-buffered saline (150 mM NaCl, pH 7.3)	24.4	456.0 ± 16.2	2.68	16.5 ± 3.1	0.31	570.4 ± 124.6	—e
Sodium acetate, 50 mM, pH 6.0	0.27	288 ± 4.8	0.11	14.8 ± 0.3	1.79	554.7 ± 10.4	0.003
Dialyzed extract (no lipopolysaccharide removal) Tris-HCl, pH 7.3	289	360 ± 4.6	66.5	12.5 ± 0.4	28.3	782.3 ± 3.4	—

^aTo compare the effect of other parameters such as buffer, salt, and the presence of lipopolysaccharide all sensorgrams were run at the same pH except for water.

^bUnit fluorescence (FU) was measured with a spectrofluorimeter (LS 50B; Perkin Elmer) with three independent measurements. Fluorescence in water and phosphate-buffered saline was highly variable, especially for native EGFP (G0).

^cThe equilibrium kinetic constant of complex lipid A-EGFP_i, K_D = k_d/k_a, was calculated from the surface plasma resonance rate constants. Association (k_a) and dissociation (k_d) rate constants were determined from experiments on E. coli lipid A monolayer immobilized on the HPA biosensor chip at 25°C with the BIACORE 2000. Kinetic constants were calculated from the sensorgrams by fitting with a 1:1 Langmuir model; all constants were derived from an average of five independent runs at five different concentrations (0.5 to 6 μM).

^dThe K_D of native EGFP (G0) was mainly attributed to the dissociation from nonspecific interaction of G0 extract.

^e—, not applicable.

Fluorescence of EGFP_i is quenched when it binds lipid A/lipopolysaccharide.

The EGFP_i displayed different fluorescence intensities at 508 nm when excited at 488 nm (Table 2). Native EGFP had the highest fluorescence, followed by G10 (62% of EGFP) and G12 (2.3% of EGFP). The drop in the intrinsic fluorescence corresponds to the increase in lysine residues that were introduced in the vicinity of the chromophore, which probably influenced the excitation and emission of EGFP. As indicated in our previous study (6), a positional effect in amino acid substitution is crucial for the intrinsic fluorescence as well its quenching during ligand interaction.

Figure 3A and B show that native EGFP exhibited a progressive fluorescence enhancement from 3.8 to 16.6% with 5 to 80 ng of lipid A and 7.7 to 29.6% with 15 to 240 ng of lipopolysaccharide. In contrast, lipid A/lipopolysaccharide binding by the G10 and G12 resulted in significant fluorescence quenching. G10 exhibited greater sensitivity to lipopolysaccharide, as the ligand induced a maximal quenching of 24.8% with 80 ng of lipid A (Fig. 3A), while 240 ng of lipopolysaccharide caused a maximal quenching of 41.4% (Fig. 3B). Removal of endogenous lipopolysaccharide also resulted in improved fluorescence quenching induced by exogenous lipid A or lipopolysaccharide.

FIG. 3.

Relative fluorescence quenching effect for 10 pmol of EGFP_i interacting with increasing concentrations of (A) lipid A and (B) lipopolysaccharide. Saturation of fluorescence quenching was observed at higher ligand concentrations that approach the limit of equal molar ratio of lipid A/lipopolysaccharide to EGFP_i, except for native EGFP (G0)_. The intensity of the fluorescent light emitted by G0 increased proportionally to the concentration of added ligand, even above the ligand-receptor ratio, exhibiting up to 16 and 29% fluorescence enhancement for lipid A and lipopolysaccharide, respectively. On the contrary, lipid A and lipopolysaccharide binding by G10 and G12 resulted in significant fluorescence quenching of these two mutant proteins. In addition, the fluorescence of G10 quenched more efficiently after lipopolysaccharide removal through the S3Δ affinity chromatography column. As more lipopolysaccharide was removed from the extracts, the quenching effect by the exogenously added ligand became more effective. The EGFP_i curves have been normalized against G0. The results are means ± standard deviation for three individual experiments. G0-S3Δ, EGFP_i G0 purified through S3Δ affinity; G10-Dial, EGFP_i G10 purified after dialysis; G12-S3Δ, EGFP_i G12 purified through S3Δ affinity; G10-TX, EGFP_i G10 purified after Triton X-14 treatment; G10-S3Δ, EGFP_i G10 purified through S3Δ affinity.

EGFP_i is a specific biosensor for live gram-negative bacteria.

EGFP_i was used to detect live bacteria because they showed high capability in binding lipid A and lipopolysaccharide in vitro. Figure 4 reveals progressive tagging of E. coli and P. aeruginosa by EGFP_i at 3, 5 and 10 min. Almost all of the cells were tagged as the reaction time reached 10 min, while the mobility of the cells decreased as incubation time was prolonged. This is consistent with our observation of the bacteriostatic effect of 0.5 to 2.0 μM EGFP_i G10 and G12, which lasted for 1 to 2 h against 10⁷ CFU of P. aeruginosa per ml (data not shown). When endogenous lipopolysaccharide was removed by purification of EGFP_i, there was improvement in fluorescent tagging of live gram-negative bacteria. On the other hand, EGFP_i did not tag S. aureus or P. pastoris.

FIG. 4.

Tagging of E. coli (A to F) and P. aeruginosa (G to I) with EGFP_i. With G10 purified by S3Δ affinity chromatography, the tagged bacteria showed increasing intensity with time of incubation from 3 to 10 min. Native EGFP (G0) was used as a negative control. P. pastoris (J to L) and S. aureus (M to O) were used as negative controls to illustrate the specificity of EGFP_i for gram-negative bacteria.

To further demonstrate the practical utility of EGFP_i, we directly challenged G10 and G12 with environmental water samples which were culture-positive for both gram-negative bacteria and gram-positive bacteria, although gram-negative bacteria predominated (Fig. 5). It is noteworthy that gram-negative bacteria were fluorescently tagged, with the exception of sheathed gram-negative bacteria. This is probably due to the additional thick layer of polysaccharide which protected the bacteria from permeation of EGFP_i proteins (Fig. 5).

FIG. 5.

EGFP_i as a biosensor for gram-negative bacteria in environmental water samples from a fish aquarium, drain water, and a stagnant pool. After a 10-min reaction time, the majority of the bacteria in the aqueous samples were illuminated by G10 (D to F) and G12 (G to I), but not by native EGFP (G0) (A to C). Gram-stained smears of the water samples show that the majority of the contaminants are gram-negative bacteria (J to L), while cultures on Mueller-Hinton agar (MHA) show bacterial colonies from the water samples: M, fish aquarium; N, drain water; O, stagnant pool.

DISCUSSION

EGFP_i were strategically designed to bind the phosphate head groups and acyl chains of the lipid A moiety of lipopolysaccharide, which is integral to the outer membrane of gram-negative bacteria. To this end, short amino acid sequences with lysine residues at a symmetrical, alternating basic and hydrophobic sequence (BHBHB binding motif) (5) were introduced into the β-sheets of the EGFP protein scaffold nearest to the chromophore. Circular dichroism analysis showed that such site-specific mutations of the EGFP scaffold, introduction of lysine residues on one or two β-strands, did not change the secondary structure of the protein (data not shown). Hence, the position of lipopolysaccharide/lipid A binding is retained in the vicinity of the chromophore in order to ensure that binding of the phosphate head groups of lipid A to EGFP_i causes fluorescence quenching (6). It was also observed from our previous study that when more lysine residues were introduced into the EGFP scaffold, the solubility and folding of the proteins were drastically affected, although the expression levels remained the same. The EGFP_i G12 mutant, which has a higher lysine content, was therefore not surprisingly expressed in inclusion bodies.

Efficient removal of lipopolysaccharide is a prerequisite to subsequent use of EGFP_i as an endotoxin biosensor. By performing at least three cycles of Triton X-114 phase separation, the endogenous endotoxin levels in recombinant proteins derived from E. coli were reduced by as much as 99%. Sequential lipopolysaccharide affinity chromatography with the Detoxi-Gel column and S3Δ affinity column (3) reduced lipopolysaccharide to <0.05 EU/ml, making it feasible for EGFP_i to bind exogenous lipopolysaccharide. We show that with each step of purification, as contaminating levels of lipopolysaccharide were progressively reduced, EGFP_i became more capable of binding exogenous lipopolysaccharide, culminating in its fluorescence quenching.

Native EGFP is acidic, pI ≈5.6. EGFP_i were engineered to contain lysine residues on one side of the β-barrel. This concentration of cationicity of the mutant proteins strengthens the electrostatic interactions with lipid A as the pH decreases. Electrostatic and hydrophobic interactions are the two major forces that synergize to increase the affinity of EGFP_i for lipid A (15). When the electrostatic interaction is weakened due to the salt, hydrophobic interaction contributes significantly to the binding. The importance of electrostatic interactions between the phosphate head groups of lipid A/lipopolysaccharide and its recognition of proteins/peptides have been well established (13). Hence, interference is envisaged from anionic groups such as phosphate in phosphate-buffered saline. Furthermore, pH and salt conditions are important factors that influence the binding of EGFP_i to lipid A/lipopolysaccharide in environmental samples.

The fluorescence of 10 pmol of G10 and G12 was quenched as more of the binding sites were occupied by the ligand (2.5 to 20 ng of lipid A or 7.5 to 60 ng of lipopolysaccharide). Above 20 ng of lipid A or 60 ng of lipopolysaccharide, the fluorescence quenching leveled off, indicating a saturating level of ligand detection. On the other hand, with EGFP, increasing levels of lipopolysaccharide/lipid A gave apparent fluorescence enhancement, although EGFP does not contain any binding sites for lipopolysaccharide or lipid A. Moreover, increasing amounts of lipopolysaccharide and lipid A did not change the K_D of EGFP, as it remained at the basal level. Therefore, one plausible explanation for the progressive rise in fluorescence with an increase in lipopolysaccharide may be that in the absence of specific binding sites in EGFP, the lipopolysaccharide acts like a “detergent” to monomerize EGFP to enhance the exposure of its chromophore, thereby increasing its fluorescence. Nevertheless, in contrast to the fluorescence quenching of G10 and G12 caused by lipopolysaccharide/lipid A binding, EGFP indeed serves as a bona fide negative control and also clearly authenticates the specific biosensing properties of EGFP_i proteins for lipopolysaccharide and lipid A.

It is clearly shown that the EGFP_i biosensor binds rapidly and specifically to gram-negative bacteria, such as E. coli and P. aeruginosa, but not other microbes, such as S. aureus and P. pastoris. In the present state, both G10 and G12 are not only capable of labeling gram-negative bacteria in laboratory cultures, they also effectively detect gram-negative bacteria in contaminated environmental samples with a certain fidelity, even in the presence of contaminants in the field samples. Perhaps some heavily sheathed bacteria with thick polysaccharide on their outer cell wall may not be labeled.

In conclusion, the present study demonstrates that EGFP_i rationally designed according to the interaction of the BHBHB binding motif to the lipid A phosphate head groups and acyl chains can be expressed in E. coli. After efficient removal of lipopolysaccharide from the lysates, EGFP_i proteins become a reliable biosensor with high binding affinity over a wide range of pHs and ionic strengths for lipopolysaccharide/lipid A and live gram-negative bacteria.